Last spring a customer sent back three lots of 11 mm moulded septa because their autosampler was throwing N=12 “no vial detected” errors on roughly 4% of injections. The septa were within the print on every dimension I could measure with a 0.01 mm micrometer. The problem was a 0.18 mm difference in the thickness at the centre versus the rim, invisible to a two-point check, and it was enough to lift the vial cap off the transport arm on a specific Shimadzu rack. That is the reality of septa dimensional tolerance capability: the number on the drawing is the easy part, and the capability of the process that produced it is the part nobody quotes.
This article is about how moulding processes actually hold dimensions on septa, how you measure the capability of a supplier rather than the capability of a single part, and which dimensions are the ones that will bite you. If you are writing an OEM specification or auditing a moulding line, this is the material that decides whether your incoming inspection is meaningful or theatre.
Septa are small, thin, and made of elastomer that shrinks unpredictably. The process you choose sets the tolerance floor before anyone touches a dial. Here is what I have measured across production runs, not what the datasheets claim.
| Process | Typical thickness tolerance (mm) | Diameter tolerance (mm) | Dimensional capability (Cpk) at ±0.05 mm | Notes |
|---|---|---|---|---|
| Compression moulding | ±0.10 to ±0.15 | ±0.15 | 0.6 to 0.9 | Flash line varies; thickness depends on charge weight |
| Transfer moulding | ±0.05 to ±0.08 | ±0.08 | 1.1 to 1.4 | Better cavity fill; gate vestige needs trimming |
| Injection moulding (L SR) | ±0.03 to ±0.05 | ±0.05 | 1.3 to 1.7 | Requires precise shot control; higher tool cost |
| Die-cut from sheet | ±0.08 (sheet gauge) | ±0.10 | 0.8 to 1.0 | Sheet gauge variation dominates; edge quality varies |
Read that table carefully. If your drawing calls for ±0.05 mm on a compression-moulded septum, you have specified a part that the process cannot make reliably, and any supplier who says yes is either not measuring or is sorting good parts out of a wide population and charging you for the scrap. I have seen both.
The ±0.05 mm figure is the practical line between “commodity” and “precision” septa. Above it, compression moulding is fine and cheap. Below it, you need transfer or injection moulding, and you need to pay for it. For a standard 11 mm GC septum the thickness spec that matters is usually 0.95 to 1.05 mm, and most reputable suppliers hold that with transfer moulding without drama. Ask them which process they use before you ask for the tolerance.
How capability is measured, and why a Cpk number on a certificate means nothing on its own
Capability is a statistical statement about a population, not a property of a part. When a supplier claims Cpk = 1.33 on thickness, the only question that matters is: on how many parts, over how many lots, and with what measurement system?
The sample size trap
A Cpk calculated from 30 parts pulled from one cavity of one lot tells you almost nothing. Elastomer moulding varies by cavity, by run, and by ambient humidity in ways that a single-lot sample will never capture. I want to see capability data from at least 25 consecutive shots per cavity, across at least 3 lots, with the cavity identified. If the supplier cannot tell you which cavity a part came from, they cannot give you a real Cpk.
Gauge R&R comes first
You cannot measure capability with a measurement system that is noisier than the tolerance. For a ±0.05 mm thickness tolerance, your gauge R&R as a percentage of tolerance should be under 10%, and under 30% at the absolute worst. A hand micrometer on a soft septum compresses the material and reads low. A non-contact optical or laser gauge is the correct instrument. If your incoming inspection uses a bench micrometer with a ratchet stop, you are measuring your own thumb pressure as much as the part.
The capability index you should actually require
For a dimension that affects fit, I specify Cpk ≥ 1.33 and Ppk ≥ 1.00 on the critical dimensions. Cpk tells you what the process is capable of when centred; Ppk tells you what it actually delivered. A supplier with Cpk 1.6 and Ppk 0.9 has a capable process that is drifting, and drifting means the next lot may not fit. That gap between Cpk and Ppk is the single most useful number on a capability report, and it is the one suppliers leave off.
Which dimensions are hard to control, and why
Not all dimensions are equal. Some are easy and some fight you on every run. Knowing which is which tells you where to put your inspection budget.
Centre thickness versus rim thickness
This is the hardest one and the one that causes the most field failures. A septum is not a flat disc; it has a raised centre or a recessed well, and the difference between the centre and the rim thickness is what determines how the cap seats and how the needle penetrates. On a compression-moulded part, charge weight variation moves the centre thickness by 0.05 to 0.10 mm shot to shot while the rim stays stable. That produces a part that passes a diameter check and a single-point thickness check and still fails in the autosampler. Measure thickness at three points: centre, mid-radius, and rim. The centre-to-rim delta should be under 0.05 mm for autosampler use.
Concentricity of the needle guide hole
If the septum has a pre-pierced hole or a guide dimple, its position relative to the outer diameter is a moulded feature and it moves with tool wear. A 0.10 mm offset is enough to change the puncture point on a 0.5 mm ID needle and start coring. This dimension is almost never on the drawing and almost always the cause when a customer reports “the needle keeps catching”.
Flash and parting-line thickness
Flash on the OD is not cosmetic. A 0.08 mm flash ring on an 11 mm septum changes the effective diameter enough to bind in a crimp cap or a screw thread. Compression moulding produces more flash than transfer, and trimming is a manual step with its own variation. Specify a maximum flash thickness and check it on the same sample as the diameter.
Coring resistance is a dimension, indirectly
The puncture count a septum survives is a function of the wall thickness around the puncture zone and the material hardness, both of which are moulded. A part that is 0.05 mm thinner in the puncture zone will core earlier. This is why two septa with identical drawings can have 40 versus 70 puncture life. The vial septa range on our sister site shows the dimensional families, but the puncture life is set on the moulding floor, not the catalogue.
Two field cases where the drawing was right and the process was wrong
Case 1: the drifting centre thickness
Symptom: a pharma QC lab running 96-well autosampler plates saw intermittent “cap not seated” alarms on one plate position, about 3% of runs, always the same position.
Measurement: we pulled 50 septa from the suspect lot and measured centre and rim thickness with a laser gauge. Centre thickness ranged 0.92 to 1.08 mm against a 1.00 ± 0.05 spec; rim thickness was tight at 0.98 to 1.02. The centre-to-rim delta hit 0.10 mm on the worst parts, which lifted the cap enough to defeat the seating sensor on that one position.
Fix: the moulder moved from volumetric charge to weight-controlled charge on the compression press and added a centre-thickness check to the in-process SPC. Centre thickness came back to 0.97 to 1.03, delta under 0.04 mm, and the alarms stopped. The drawing never changed. The process did.
Case 2: the concentricity offset
Symptom: an environmental lab running EPA Method 8260 on a GC-MS reported coring after 30 to 35 injections on a septum rated for 100. They blamed the needle.
Measurement: we sectioned 20 septa and measured the guide dimple position relative to the OD. Mean offset was 0.14 mm, with a range of 0.08 to 0.21. The tool had worn on one side of the cavity. The needle was entering off-centre and tearing rather than piercing.
Fix: tool refurbishment and a concentricity check at 0.08 mm TIR added to the drawing. Puncture life returned to 90+ injections. The SW-846 method never mentions the septum, but the data quality did.
The tolerance stack that decides whether your autosampler works
No septum dimension matters in isolation. What matters is the stack: septum thickness, cap crimp depth, vial neck height, and autosampler rack position. A septum that is 0.05 mm over nominal is fine if the cap crimp is 0.05 mm under. It is not fine if both drift the same way.
For a standard 11 mm crimp vial, the working stack is:
- Vial neck height: ±0.10 mm (glass, usually the tightest controlled)
- Septum thickness: ±0.05 mm (moulded, the variable one)
- Crimp cap skirt depth: ±0.08 mm (metal, tool-dependent)
- Crimp force: 1.5 to 2.0 kN on a manual crimper, and it varies with operator
The septum is the only component in that stack that is soft, and soft components absorb variation by deforming, which is why a bad septum sometimes works and a good one sometimes does not. The stack analysis is what tells you whether your ±0.05 mm spec is enough. If the total stack exceeds the autosampler’s seating window (typically 0.15 to 0.20 mm), you have a problem no single-component tolerance will solve. The autosampler vial geometry sets the window; the septum has to live inside it.
This is also why I ask for the stack drawing, not just the septum drawing, when I qualify a new moulding source. A supplier who has never seen the cap and vial they are mating to cannot know which dimension to prioritise. The HPLC vial and cap families on our sister site are built to a common stack for exactly this reason.
What to put on the drawing, and what to leave off
A septum drawing with 30 dimensions and no capability statement is a wish list. Here is the decision checklist I use when specifying a moulded septum for an OEM or a regulated lab.
- Thickness: nominal ± 0.05 mm, measured at centre and rim, with a maximum centre-to-rim delta of 0.05 mm. This is the dimension that decides fit.
- Outer diameter: nominal ± 0.08 mm, with a maximum flash thickness of 0.05 mm on the OD.
- Concentricity: 0.08 mm TIR for any guide feature. Non-negotiable for autosampler use.
- Hardness: Shore A ±3 within a lot, ±5 across lots. Hardness drift changes puncture life more than any other single variable.
- Capability: Cpk ≥ 1.33 and Ppk ≥ 1.00 on thickness, OD, and concentricity, reported per cavity across at least 3 lots.
- Material: silicone or EPDM to a named durometer and a named cure system, with a certificate of analysis per lot.
- Cleanliness: no mould release agent on the puncture face, and a stated extractables limit if the application is trace analysis.
Leave off the dimensions that do not affect function. A cosmetic flash line on the underside, a surface finish number you cannot verify, a colour match to a Pantone chip: these add cost and inspection time and buy you nothing. Every dimension you put on the drawing is a dimension someone has to measure, and the more you add, the more likely the important ones get skipped.
For OEM programmes, the drawing should also name the process (transfer or injection moulding) and the cavity count. A 16-cavity tool and a 64-cavity tool do not have the same capability, and a supplier who runs both will not tell you which one made your parts unless you ask. The septa we mould for OEM customers are quoted with the cavity count and the per-cavity capability on the first page of the PPAP, because that is the number the customer’s quality engineer actually needs.
How to audit a moulding line in one afternoon
You do not need a three-day audit to find out whether a supplier can hold tolerance. You need four things.
First, ask to see the in-process SPC chart for the last 30 days. Not the certificate, the chart. If thickness is plotted with control limits and the last 30 points are inside them, the process is in control. If the chart shows a trend or a step, ask what changed. If there is no chart, the supplier is inspecting output, not controlling process, and you are buying sorted parts.
Second, ask for the gauge R&R on the thickness measurement. If it is over 30% of tolerance, their capability numbers are noise. I have walked away from suppliers over this alone.
Third, pull 10 parts from three different cavities and measure them yourself. A laser gauge or an optical comparator will do. If the cavity-to-cavity spread is more than half the tolerance, the tool is worn or the process is not balanced, and no certificate will save you.
Fourth, check the material certificate against the durometer on the drawing. A lot of “tolerance failures” are actually material substitution. A batch of EPDM at Shore A 45 instead of 40 will read 0.03 mm thicker at the same compression and core 20 injections earlier. The Agilent consumables and Thermo Fisher sample handling catalogues both publish durometer and thickness for their septa, which gives you a benchmark to hold a new supplier against. Restek publishes puncture-life data alongside dimensions, which is the closest thing to a capability statement the industry offers publicly.
The decision rule for specifying tolerance
Specify the tightest tolerance the mating stack requires, then verify the supplier’s process can hold it at Cpk ≥ 1.33 per cavity, then stop. A tighter tolerance than the stack needs costs money and buys nothing; a looser one than the stack needs guarantees field failures that no incoming inspection will catch, because the failures are in the interaction, not the part. If a supplier cannot show you per-cavity capability data across multiple lots with a gauge R&R under 30%, the correct decision is not to negotiate the tolerance. It is to find a supplier who measures.
References & further reading
Related resources from our lab consumables network:
- Septa at septacaps.com
- Vial Septa at vialsepta.com
- Hplc Vial at hplcgc.com
- Autosampler Vial at autosamplerhplc.com
External references and standards:
